1

The thing I like about the Google Speech to Text API, is that it has fantastic accuracy, a fair bit better than the Azure API in my experience. The reason being, I am able to send a list of expected words in my request to the Google API, and therefore, increasing the likelihood of those terms/phrases occuring. With Azure, I am restricted to making speech models. The problem with this is, I would have to create a new model for each request, as each request will have a different vocabulary etc. from the rest. The requests are not industry specific, they are completely random.

So, my question is, can I send a list of expected words with my request to the Azure API?

Cheers

Harry Stuart
  • 1,781
  • 2
  • 24
  • 39

1 Answers1

0

have you solved your problem?

Concerning Azure, you need to first gather training data that is relevant to your problem and industry. You then train one model on this dataset, and in production you use this model to transcribe your audio.

You don't train one model every time you have a new request.

Be Chiller Too
  • 2,502
  • 2
  • 16
  • 42